Lecture no 18: Redundancy

1
Lecture no 18 Redundancy
TDT4285 Planlegging og drift av
IT-systemer Spring 2007 Anders Christensen, IDI
2
Definition

• Redundancy is using extra, superfluous capacity
  to establish a safety margin against faults in
  the system.

3
Types of redundancy

• Fail-over seemless transition to another unit
  providing identical service
• Backup An extra copy of important data.
• Duplication Several units provide the same
  services, and they can cover up for eachother if
  one of them fails.
• Fall-back alternative, but usually less
  attractive solution, often temporary

4
Fail-over and full redundancy
server1
The principle is the same as for RAID
Originally
Two or more equivalent servers
client
Auto fail over
mirroring
server2
After fault on server1
5
Over-capacity and N1redundancy
server1

• Needs three units
• Has four units
• Extra capacity is a useful side effect
• By faults, it is possible to manage with a
  reduced capacity.

server2
server3
server4
Extra unit
6
Single point of failure

• Definition SPOF is a component which the
  totality relies on.
• Examples Power, computer networks
• Redundancy eliminate SPOF.

spof
spof
7
Akkumulation of down-time
80

• Down-time or a system consisting of several
  subunits depends on the down-time of these units.
  
• This is the calculation of probability.
• Note that there is a difference between units
  that fail independant of eachother, and where
  they feil synchronously.

90
50
50
90
100
8
SPOF may also be what you need...

• Allows you to control where a fault happens (like
  in a fuse).
• Allows you to detect and have overview over
  faults.
• The use is simultaneously a testing
• Its cheaper (unless you need the up-time and
  quality)

9
Example RAID5
Note simplified model
Data
Parity
Extra
Disk1
Disk2
Disk3
Disk4
Disk5
Disk6
Disk7
rebuild of disk2
10
Full and N1redundancy

• N1-redundancy
• One extra unit of every component of a certain
  type
• Usually HW
• Used for selected parts of the system

• Full redundancy
• The whole system is duplicated
• Expensive (gt2cost)
• May be implemented for a whole site.

11
Redundancy traps.

• Fail-over becomes double life-time
• Identical units with identical bugs
• Installed but un-tested
• Bought but false safty
• No guarantee against human errors
• Can become a white elefant

12
Example HVAC at IDIs computer room
Fan
Power
Duplicated unit at other side of room
Closed loop cooling fluid
Radiator
Computer floor
Water supply
13
Costs

• Reduncy costs more (in operations!)

Extra equipment
Higher complexity
Note only approximately calculation
More maintenance
- Advantage of higher up-tim
Extra costs
14
Costs vs up-time
99,9
Costs
99
90
Up-time
Alternative strategy recovery in stead of
redundancy
15
Return On Investment (ROI)

• Result with current solution (R1)
• Result with investment (R2)
• Cost of investment (Invest)
• ROI (R2 R1) Invest

16
Example A car

• Tires 4spare tire (N1 redundancy)
• Petrol Spare petrol tank (backup?)
• Break lights 2-3 units (full redundancy)
• Breaks hand break (fall-back)
• Taxi Can call for (out-soucing)
• Safetybelt and collision bag different
  implementation, common use (duplication)
• Head lights can use parking lights (fall-back)